Neural network modeling for small DMFC bipolar plate stack system

Since the multitudinous parameters make the DMFC model become a complex "black-box" direct methanol fuel cell (DMFC) modeling has become an important and difficult issue of practical experiment. This paper presents the model of the small DMFC Bipolar Plate Stack System. This system is different from the conventional complex mathematical models by the neural network (NN) which is considered the real paramneters of the measurement inputs and outputs. First the training data is the power density of short-term stability tests for the DMFC 10-cell stacks. Then the DMFC model can be obtained by NN method and contented to the different unit cell stacks. Finally, the simulation results in agreement with experimnental results show that the NN modeling method effectively projects the power density on small DMFC packs, such as the developmnent and simulation tool. Therefore the NN modeling method can save much time to reform the conventional mathematical models by the very expensive experimnent.